A computerized approach to glycemic control in critical illness
Authors: Michael A. Grasso
Book Title: American College of Physicians, Maryland Regional Associates Meeting, Baltimore Maryland
Date: May 30, 2008
Abstract: Background: Hyperglycemia is a known problem in the critical care setting. Impaired compensatory response to β cells and increased counter-regulatory hormones are thought to stimulate endogenous glucose production and contribute to insulin resistance. A growing body of evidence advocates strict maintenance of normoglycemia in order to decrease morbidity and mortality. The American College of Endocrinology recommends a blood glucose upper limit of 110 mg/dL for ICU patients and 180 mg/dL for patients outside the ICU. Standardized protocols for subcutaneous and intravenous insulin have been developed. However, for a variety of reasons, guidelines are not always adhered to. Several groups have attempted to computerize these protocols in order to improve compliance. The objective of this study was to perform a literature review of computer-based insulin protocols in the intensive care setting.
Methods: A PubMed review was conducted using the keywords (insulin or diabetes mellitus) and (software or software validation or software design or computer) and (intensive care or hospitalization or inpatients) in February 2008.
Results: The search returned 80 articles published between 1971 and 2008. Fifteen of these articles, published between 2002 and 2008, discussed computer-based insulin protocols in the intensive care setting. The majority of these efforts used software that was independent of existing electronic medical record systems, required duplicate entry of data already found within the medical record, did not provide a rationale with each recommendation, and used little historical data to customize a patient's insulin regimen.
Conclusions: Based on lessons learned from these efforts, an optimal system would incorporate the following characteristics. It would fit into the existing clinical workflow through a tight integration with computerized physician order-entry systems. To avoid duplicate data entry, it would automatically extract data from the electronic medical record or glucometers. To increase confidence and encourage compliance, it would provide an explanation with each recommendation. Finally, it would take advantage of the computational model by incorporating complex algorithms and historical data when making recommendations.
Google Scholar: search